Evolution

Two facts about living things demand an account: they fit their circumstances with an accuracy that looks designed, and they fall into a nested pattern of similarity that design has no reason to produce.

Whether a slow mechanism is available at all is a question for geology and physics, and the answers arrived in the wrong order: extinction first, an ancient earth second, and the numbers only in the twentieth century.

The Origin of Species is one long argument from four premises nobody disputed to a conclusion almost nobody accepted, and its weakest joint is the one Darwin knew least about.

If inheritance blends, natural selection destroys the variation it feeds on within a few generations, and the experiment that solved the problem had already been published and ignored.

Mendelian heredity conserves variation by itself, which turns evolution from a story about individuals into a measurable quantity: the frequency of an allele in a population.

Break one Hardy-Weinberg assumption and the population starts to move, at a rate the algebra predicts and two field cases let you check against real frequencies.

Selection is a filter and cannot make what it sorts, so the theory needs a source of new alleles: a measurable rate of copying error that arrives before the challenge rather than in response to it.

In a finite population allele frequencies wander whether or not anything is selecting them, which decides the fate of most new mutations and predicts a molecular clock.

A lineage that only changes gives one species at a time, so diversity requires splitting, and splitting means the evolution of barriers to interbreeding.

If lineages split and rarely rejoin, the history of life is a tree, and that is a testable claim because thousands of independent characters must agree on which tree it is.

The record is spectacularly incomplete, which is why a fossil found by predicting where it must lie is worth more than any number found by luck.

Every organism carries a complete record of its ancestry, and the decisive entries in it are not the working parts but the shared mistakes, which common design has no reason to copy.

The process is fast enough to measure while somebody is watching, which turns the theory from a reconstruction of the past into a quantitative tool with a bill attached.

Constraint, history and chance bound what selection can build, the standing objections resolve into the theory rather than against it, and several real arguments about it are still open.